I was wondering if anybody has a more efficient, more intelligent way to do this. The looping construct requires every deleted files to be expunged from every commit by reading each commit. With many commits, this takes a long time.
git log --pretty=format: --name-status | grep -i ^D | cut -f2- | sort -u | xargs -I {} git filter-branch -f --index-filter 'git rm --cached --ignore-unmatch {}' HEAD That seems to remove each files one at a time.
Consider that git rm can take multiple files to be removed.
So one optimization would be to build that list and call the filter-branch once.
You can see one example of that approach in "Proper way to remove unwanted files with git filter-branch without git rm failing".
grep -v -x -f /tmp/files-in-repo.txt along with xargs --delimiter=\\n -r git rm ... useful.Just a small improvement to support recursive delete
git filter-branch --prune-empty --index-filter 'git log --pretty=format: --name-status | grep -i ^D | cut -f2- | sort -u | xargs git rm -r --cached --ignore-unmatch DoesNotExistInMyProject' For anyone else who might be looking for the same answer, I modified the original command above with the suggestions linked in the other thread and created this command which seemed to work for me.
git filter-branch --prune-empty --index-filter 'git log --pretty=format: --name-status | grep -i ^D | cut -f2- | sort -u | xargs git rm --cached --ignore-unmatch DoesNotExistInMyProject' Improved the command to avoid file names with a ' which causes xargs to bug
echo "/!\ This file contains a single quote ' and it's gonna bug with xargs, so it's excluded from the filter operation" ; bash -c "git log --pretty=format: --name-status | grep -i ^D " | grep \' git filter-branch -f --prune-empty --index-filter "git log --pretty=format: --name-status | grep -i ^D | grep -v \' | cut -f2- | sort -u | xargs git rm -r --cached --ignore-unmatch DoesNotExistInMyProject"
